Click on "Install Server".
Wait a few minutes for the server to deploy. Once ready, it will show a "Started" state.
In the chat, type @ followed by the MCP server name and your instructions, e.g., "@ContractAudit MCP Server search for reentrancy vulnerabilities in ERC20 tokens"
That's it! The server will respond to your query, and you can continue using it as needed.
Here is a step-by-step guide with screenshots.
ContractAudit is an audit research assistant: it crawls an explicit source allowlist, extracts HTML/PDF, builds provenance-preserving chunks (LlamaIndex), stores hybrid dense + BM25 vectors in Qdrant, and exposes read-only search through MCP.
It helps you find what auditors and docs already said about a class of bugs. It is not a certificate that any contract is safe.
approved sources → governed crawler → parsers → LlamaIndex chunks
→ Qdrant (hybrid) → retrieval service → MCP tools → (optional) local LLM hostCapability | Details |
Governed crawling | Domain / path allowlists, crawl delay, size caps, |
Ingestion | HTML + PDF extraction, chunking with stable IDs + provenance |
Hybrid search | Dense embeddings ( |
MCP server | Read-only tools for IDE / agent hosts ( |
Eval harness | Starter benchmark queries in |
LLM seam | Optional host retrieves evidence via MCP, then calls your local |
MCP tools: search_security_knowledge · get_audit_finding · get_document_context · list_sources · corpus_status
Crawling and indexing stay operator-controlled CLI actions so prompt content cannot mutate the corpus.
Related MCP server: claudit
Layer | Technology |
Language | Python 3.11+, packaged with Hatchling |
CLI | Typer ( |
Config | Pydantic Settings, YAML source policies |
Crawl / parse | httpx, BeautifulSoup, trafilatura, pypdf |
Chunking / RAG | LlamaIndex + HuggingFace embeddings |
Vector DB | Qdrant (embedded path or server URL) |
Sparse vectors | fastembed |
Agent interface | MCP ( |
Quality | pytest, ruff, mypy (strict) |
Optional: OCR extras (pymupdf, pytesseract) · docs PDF builder (reportlab).
Included | Excluded (local only) |
Source code, tests, |
|
|
|
Learning guide (md/pdf) |
|
No API keys are required for the default local embedding path. Do not commit crawled corpora or vector stores.
python -m venv .venv
.\.venv\Scripts\Activate.ps1
python -m pip install --upgrade pip
python -m pip install -e ".[dev]"
Copy-Item .env.example .envFirst search/ingest downloads embedding models. Default dense model runs on CPU; set CAR_EMBEDDING_DEVICE=cuda only if VRAM allows.
Embedded Qdrant (data/qdrant) allows one process at a time. For concurrent ingest + MCP, run Qdrant as a service and set CAR_QDRANT_URL=http://localhost:6333.
Review config/sources.yaml first (robots, terms, report licenses).
contract-audit-rag sources validate
contract-audit-rag crawl --source trailofbits_secure_contracts --limit 30
contract-audit-rag ingest
contract-audit-rag stats
contract-audit-rag search "How should oracle price freshness be checked?"
contract-audit-rag benchmarkcontract-audit-mcpFor a local network client: CAR_MCP_TRANSPORT=streamable-http (default 127.0.0.1:8765). Do not expose publicly without auth/TLS.
ruff check .
mypy src
pytestWire any local model callable through contract_audit_rag.llm.base.CallableAdapter, then use MCPQwenHost.answer() with a connected MCP ClientSession. The host:
Calls search_security_knowledge
Validates structured evidence
Builds an evidence_prompt (untrusted web content, required citations, insufficient-evidence path)
The raw model is not an MCP client — the application host owns tool calls.
ContractAudit/
├── config/sources.yaml # Crawl allowlist (review before use)
├── src/contract_audit_rag/
│ ├── cli.py # Typer CLI
│ ├── ingestion/ # Crawler, parsers, pipeline, chunking
│ ├── retrieval/ # Search service
│ ├── indexing.py # Qdrant index store
│ ├── mcp/server.py # MCP tools
│ └── llm/ # Optional host + adapter seam
├── eval/evm_queries.yaml
├── tests/
├── docs/ # Learning guide (md + pdf)
├── tools/build_learning_guide.py
├── .env.example
└── pyproject.tomlAllowlist-first security posture for anything that hits the network.
Provenance-preserving chunks so answers can be cited, not hand-waved.
Read-only MCP surface — corpus mutation is never a tool side effect.
Hybrid retrieval for both semantic and keyword-heavy audit jargon.
Honest product boundary: research assistant ≠ automated audit sign-off.
Detailed walkthrough: docs/Contract_Audit_RAG_Learning_Guide.pdf (Markdown source alongside).
python -m pip install -e ".[docs]"
python tools\build_learning_guide.pyMIT — see LICENSE. Respect third-party content licenses when crawling or redistributing reports.
Return corpus, vector collection, and embedding model status.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
| Name | Required | Description |
|---|---|---|
No output parameters | ||
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations provided; description carries full burden. Only states it returns status but does not disclose any behavioral traits like read-only nature, authentication requirements, or side effects.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
Single, concise sentence that directly conveys the tool's function with no excess words.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given zero parameters and presence of output schema, description is adequate but could mention if status is real-time or any prerequisites. Siblings provide enough differentiation.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
No parameters (coverage 100%), baseline 4. Description adds meaning by specifying the components of the status (corpus, vector collection, embedding model), which is helpful beyond the empty schema.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
Clearly states verb 'Return' and resource 'corpus, vector collection, and embedding model status'. Differentiates from sibling tools get_document_context and list_sources, which have different purposes.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
Implies usage for obtaining status but does not explicitly state when to use or when not to use. No mention of alternatives or exclusions.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Return ordered chunks from one source document.
| Name | Required | Description | Default |
|---|---|---|---|
| limit | No | ||
| document_id | Yes |
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
With no annotations provided, the description carries full burden but only states 'Return ordered chunks,' implying a read operation. No behavioral traits (e.g., permissions, rate limits, idempotency) are disclosed, which is insufficient for a mutation-free analysis tool.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
The description is a single sentence with no waste, but it is too brief for a tool with parameters and no schema descriptions. Front-loading the core action is good, but essential details are missing, making it minimally adequate.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given the presence of an output schema, return values are not strictly required. However, the description fails to explain how the 'ordered chunks' are organized, how 'limit' affects results, or what 'document_id' refers to, leaving significant gaps for a tool with two parameters and no param descriptions.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
Schema description coverage is 0%, yet the description adds no meaning to the two parameters. 'document_id' and 'limit' are not explained, leaving the agent to infer their purposes without any additional context.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the tool returns ordered chunks from one source document, specifying the verb (Return) and resource (ordered chunks from one source document). It distinguishes from sibling tools like list_sources and corpus_status, which likely provide different functionality.
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance is provided on when to use this tool versus alternatives. There is no mention of prerequisites, limitations, or exclusion criteria, leaving the agent without context for appropriate usage.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
List ingested source documents, licenses, URLs, and processing states.
| Name | Required | Description | Default |
|---|---|---|---|
No parameters | |||
| Name | Required | Description |
|---|---|---|
| result | Yes |
Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?
No annotations are provided, so the description carries the burden. It indicates a read operation ('list'), which is inherently safe, but does not disclose any additional traits like pagination, rate limits, or behavior when empty.
Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.
Is the description appropriately sized, front-loaded, and free of redundancy?
A single, front-loaded sentence with no extraneous words. Every word earns its place.
Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.
Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?
Given no parameters and an existing output schema, the description adequately covers the basic purpose. It could mention scope (all sources) but is sufficient for a simple list operation.
Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.
Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?
There are zero parameters, so baseline is 4. The description can add value but is not required to explain parameters.
Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.
Does the description clearly state what the tool does and how it differs from similar tools?
The description clearly states the verb 'list' and the resource 'ingested source documents, licenses, URLs, and processing states', and it distinguishes from siblings like get_document_context (single document) and corpus_status (overall status).
Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.
Does the description explain when to use this tool, when not to, or what alternatives exist?
No guidance on when to use this tool vs alternatives or any exclusions. The description only states what it does without context.
Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.
Each tool targets a clearly distinct aspect of the server: retrieving document content, listing source metadata, and reporting system status. No overlapping functionality.
Most names follow a verb_noun pattern (get_document_context, list_sources), but corpus_status is a noun_noun pair without a verb, breaking full consistency. Overall pattern is clear and readable.
Three tools is on the low end for a contract audit server, but could be acceptable if the server is narrowly scoped. It borders on feeling thin.
The tool set lacks essential operations like adding or deleting sources, searching across documents, or comparing contracts. Significant gaps impair full audit workflows.
Unclaimed servers have limited discoverability.
If you are the server author, to access and configure the admin panel.
Evidence-backed crypto due diligence with sources, freshness, and a runtime receipt on every call.
Securely search and manage workspace context files for AI agents and teams.
Read-only semantic search over Vedic scripture verses, commentaries, and recorded lectures.
Read-only search of your Sortio knowledge graph (files and entities) for Claude and ChatGPT.
We provide all the information about MCP servers via our MCP API.
curl -X GET 'https://glama.ai/api/mcp/v1/servers/ArmaghanRazaChaudhary/ContractAudit'
If you have feedback or need assistance with the MCP directory API, please join our Discord server
